[Coin-standards] INFORMS / directions

[S. Lozano]

Dear colleagues,

My name is Sebastian Lozano. I am professor here at the University of 
Seville. Presently I am engaged in a research work together with a Ph.D. 
student on trying to define an XML application for mathematical 
modeling. That is why, when I found Coin-standards a couple of weeks ago 
I subscribed immediately and I am very interested in the subject. Our 
research goals are rather ambitious and perhaps you might help us to 
assess them better.

Our rationale is that, no doubt algebraic modeling languages are 
powerful and convenient for OR practitioners and they have a loyal 
following. However, there are mutiple modeling languages (GAMS, AMPL, 
LPL, AIMMS, MPL, etc) each one with its own particular sintax. Although 
these environments allow for interactive use, we will concentrate in its 
static use, i.e. producing a model file and a data file with specific 
keywords and separators, etc. It seems to us that codifying the model as 
well as its data in XML may have benefits in terms of standardisation, 
easiness of validation, portability, etc. To that end we are in the 
process of analysing the different modeling languages in order to find 
their common subset of functionality and devise an XML Schema for an 
XML-based modeling language. Since we do not pretend neither to compete 
nor to discourage the usage of the established language (but to 
complement them) we would like to keep the possibility of transforming 
(through XSL stylesheets) the problems expressed using  XML the existing 
modeling languages. The opposite (i.e.  being able to parse files 
expressed in existing modeling languages and generate the corresponding 
XML representation) although desirable would only be feasible for 
certain types of models but not in general.

With respect to the question posed by some of you in previous messages, 
we first considered the possibility of just expressing in XML something 
like the MPS or SMPS formats. Although that would be feasible, we 
thought that a better approach would be to imitate modeling languages 
since that allows a more compact representation of problems and would 
allow the separation of models (whose XML representation would be 
reusable) from data specific from each instance. However, that does not 
preclude the possibility of transforming (again using XSL) the compact, 
algebraic XML representation  into a more verbose MPS-equivalent XML 
representation.

About saving the XML model and data files in native databases and 
accessing them using QPath the thruth is that we have not studied that. 
We did consider the use of MathML but for the modeling phase. For the 
moment we have discarded it and prefer to follow the existing (and 
succesful) modeling languages path instead.

As I mentioned above, a Ph. D. student is working on this for his 
doctorate but our aim is to collaborate with similar efforts in the OR 
community so that whatever comes out be shared and supported as widely 
as possible since otherwise it would only be an academica exercise.

Thank you for your patience, yours sincerely

-- 
Sebastian Lozano
Escuela Superior de Ingenieros
Camino de los Descubrimientos, s/n 
Isla de la Cartuja
E-41092 Sevilla (Spain)

Phone:  +34-95448-7208
Fax:    +34-95448-7329 / +34-95446-3153